Login at Kodi Home

maxk · 2016-12-22, 20:01

(2016-12-22, 16:54)natethomas Wrote: Are you advocating killing the Android port or advocating fixing the windowing system across platforms including the android port?

The latter, of course!

The code is bad, everybody agrees on that. But we don't agree where to go from here.

Team Kodi decided to delete one random bad windowing system (Wayland), but didn't delete all the others which share the same (bad) design, including Android.

My suggestion was to get Wayland back in, and then refactor all EGL-based windowing systems at the same time (i.e. Wayland, Android and the others). That was not only a suggestion - I volunteered to do it, and I already got it started, but it was rejected. Even though it was just like fritsch suggested in the very same thread, the same fritsch who rejected his own idea in this forum.

Today, a year later, Kodi is still without Wayland support.

**Martijn** · (This post was last modified: 2016-12-22, 20:04 by Martijn.)

our suggestion was to refactor it first and then get it in. See the difference here?

**fritsch** · 2016-12-22, 20:05

Interpreting things is freedom of speech. Therefore thanks for your interpretation.

maxk · 2016-12-22, 20:12

(2016-12-22, 20:03)Martijn Wrote: our suggestion was to refactor it first and then get it in. See the difference here?

Your memory doesn't appear to serve you well. I know both of that already, because I explained it to you already:
Explanation 1, Explanation 2

**Martijn** · (This post was last modified: 2016-12-22, 20:16 by Martijn.)

Nothing wrong with my memory. You explained but explanation was not accepted as valid.

(2016-12-22, 20:05)fritsch Wrote: Interpreting things is freedom of speech. Therefore thanks for your interpretation.

Looks like i can quite that as well for this Smile

qp9013625 · 2016-12-22, 21:13

(2016-02-01, 16:43)FernetMenta Wrote: - no support for VDPAU (used by AMD and NVIdia)

Wayland supports VA-API. So couldn't hardware acceleration on AMD and NVIDIA be achieved by using libva-vdpau-driver:

https://cgit.freedesktop.org/vaapi/vdpau-driver/

?

**fritsch** · 2016-12-22, 21:19

No. As both don't support render nodes via vaapi and additionally vdpau is tied to X. Wayland will be the definite dead for vdpau as of now.

Besides that the proposed driver uses opengl internally to trnansition vaapi input to vdpau interface ... which - you guessed also won't work.

**natethomas** · 2016-12-23, 03:52

@maxk, I'm reading through that post, and I'm a bit at a loss as to why the need to merge before fixing everything. If you are willing to work with fernet, couldn't the two of you create a separate branch, merge in your "broken" Wayland (not trying to offend, just trying to use the terminology I read in the PR), and then proceed to fix the windowing system across platforms? If you do that, and the team knows, I imagine people would try to avoid touching the current system in master, to reduce rebasing problems in the future.

FernetMenta · 2016-12-23, 08:47

(2016-12-22, 19:46)qp9013625 Wrote: Or are you just saying that the patch that adds Mir support to KODI is better than the patch that added Wayland support to KODI?

Exactly. BTW this work was done by a developer from Canonical. It was right on first attempt. That tells us that the architecture how a windowing system needs to be implemented in Kodi is rather obvious. The dependencies to other parts of the application are kept to a minimum. That also means that work on this component does not interfere with work done in other areas.

X11 is dead, no doubt, and Kodi needs to support Wayland. The way how this needs to be done is clear: do it the same way Mir support was done.

maxk · 2016-12-23, 09:48

(2016-12-23, 03:52)natethomas Wrote: @maxk, I'm reading through that post, and I'm a bit at a loss as to why the need to merge before fixing everything. If you are willing to work with fernet, couldn't the two of you create a separate branch, merge in your "broken" Wayland (not trying to offend, just trying to use the terminology I read in the PR), and then proceed to fix the windowing system across platforms?

Because then I would be working on the code for a year or so, without any guarantee that my code would ever be merged. There isn't even the most basic agreement on how the new architecture should look like - I asked several times, but never got a reply. This isn't going anywhere.
I would be forced to do the rewrite alone. If it's not in mainline, nobody will help.
With (bad) Wayland back in mainline, everybody can benefit from the small incremental improvements being made to all windowing systems, every day. With every pull request from me, everybody could take part in the discussion and steer the new architecture in the right direction, instead of rejecting my overall solution in the very end, after I invested thousands of hours into it.

There is nothing to lose with my approach - it can't go wrong, because it's a community effort, and fine-grained checks and balances are there. But there's a lot to be won.

On the other hand, with your approach, there is much to lose, and only small chances for a win.

Where: "much to lose" = thousands of hours of my time. Since the loss is only on my side, it's easy for you to demand this approach, but that's selfish.

**natethomas** · 2016-12-23, 11:08

(2016-12-23, 09:48)maxk Wrote: Where: "much to lose" = thousands of hours of my time. Since the loss is only on my side, it's easy for you to demand this approach, but that's selfish.

You are putting words into my mouth if you think what I said was a demand. You obviously want to work on this project, and I'm trying to think of a way to make this happen that would make everyone happy. Perhaps you are prepared for people being combative, so you automatically assume every reply is combative, or maybe being combative is just part of your nature, but for the sake of this conversation, I'd like to request that you at least start with the assumption that people aren't out to attack you or make demands of you.

maxk · 2016-12-23, 11:28

(2016-12-23, 11:08)natethomas Wrote: You are putting words into my mouth if you think what I said was a demand.

I don't. That's not what I wrote, and not what I meant.

Quote:I'm trying to think of a way to make this happen that would make everyone happy.

It's certainly a good idea to think of alternative approaches if we're stuck (and indeed we are, but you just paraphrased what was already suggested). I was just arguing why your suggestion is inferior, and why I consider the Team's rejection of my suggested approach a bad choice. So bad that I fear (after many code rejections for stupid reasons) that all the time I'd invest would be wasted.

FernetMenta · 2016-12-23, 19:00

(2016-12-23, 09:48)maxk Wrote: Where: "much to lose" = thousands of hours of my time. Since the loss is only on my side, it's easy for you to demand this approach, but that's selfish.

Well, if you think you need that long, you really should look for something different you can do in your spare time. Note that the first implementation for Wayland and the Mir implementation we just integrated were done by single persons without the guarantee that the work will be taken by the project.

maxk · 2016-12-23, 19:04

(2016-12-23, 19:00)FernetMenta Wrote: Well, if you think you need that long, you really should ...

Point missed, by a large margin.

**natethomas** · 2016-12-23, 21:17

I think you missed a fairly critical part of my suggestion. You are worried that the resulting PR wouldn't be accepted, but my suggestion was that you do the work with in that separate thread with the very people who will be doing the accepting, so you know with a fair degree of certainty that it would be accepted.